Probabilistic Rapid Intensification (RI) Forecasts in the Atlantic and E. Pacific Basins

Methodology:

The OAR has identified obtaining guidance on the timing and magnitude of episodes of rapid intensification (RI) as one of the highest operational tropical cyclone forecasting priorities. In an effort to provide a tool to aid in the forecasting of rapid intensification, an updated index for estimating the probability of rapid intensification (Kaplan and DeMaria 2003) has been developed as part of the NOAA Joint Hurricane Testbed. The rapid intensification index employs 5 large-scale and 2 satellite-derived inner-core predictors from the SHIPS model (DeMaria et al., 2005) to estimate the probability of rapid intensification for the 24-h period commencing at t=0. The 2006 version of the rapid intensification index differed from that used in previous seasons in a few areas. First, the threshold for rapid intensification was lowered slightly from a maximum sustained wind increase threshold of = 30 kt in 24 h to a threshold of = 25 kt in 24 h. Also, the methodology for deriving the index was modified so that that the contributions from each of the rapid intensification predictors represented a scaled value between 0 and 1 rather than simply a 0 or 1 as had been employed in the previous version of the index. Finally, a second version of the RI index was derived using predictor values that were weighted utilizing linear discriminant analysis (Wilks 1995). A more complete description of the methodology used to derive the 2006 rapid intensification index can be found in Kaplan and DeMaria (2006).

Principal Investigator:

John Kaplan

Collaborating scientist:

Mark DeMaria (NOAA/NESDIS/ORA)

Methodology:

The OAR has identified obtaining guidance on the timing and magnitude of episodes of rapid intensification (RI) as one of the highest operational tropical cyclone forecasting priorities. In an effort to provide a tool to aid in the forecasting of rapid intensification, an updated index for estimating the probability of rapid intensification (Kaplan and DeMaria 2003) has been developed as part of the NOAA Joint Hurricane Testbed. The rapid intensification index employs 5 large-scale and 2 satellite-derived inner-core predictors from the SHIPS model (DeMaria et al., 2005) to estimate the probability of rapid intensification for the 24-h period commencing at t=0. The 2006 version of the rapid intensification index differed from that used in previous seasons in a few areas. First, the threshold for rapid intensification was lowered slightly from a maximum sustained wind increase threshold of = 30 kt in 24 h to a threshold of = 25 kt in 24 h. Also, the methodology for deriving the index was modified so that that the contributions from each of the rapid intensification predictors represented a scaled value between 0 and 1 rather than simply a 0 or 1 as had been employed in the previous version of the index. Finally, a second version of the RI index was derived using predictor values that were weighted utilizing linear discriminant analysis (Wilks 1995). A more complete description of the methodology used to derive the 2006 rapid intensification index can be found in Kaplan and DeMaria (2006).

Results:

Figure 1 shows the skill of the RI index for the 1995-2005 Atlantic and E. Pacific dependent samples. The skill shown in Fig. 1 was computed relative to climatology using brier skill scores where positive values indicate skill relative to climatology and negative values denote skill that is less than that of climatology. The figure indicates that the E. Pacific version of the index exhibits more skill than the Atlantic version and that the discriminant version of the index outperforms the scaled version. Figure 2 shows the skill of the independent RI index forecasts that were provided to the OAR in real-time during the 2006 hurricane. For comparison, the figure also shows the skill of the other operational deterministic intensity guidance as well as the official NHC forecast. To compute the skill of each of the deterministic models and the official NHC forecasts it was necessary to assign a probability to each forecast. This was accomplished by assigning either a probability of 100% or 0% to each case based upon the forecasted intensity change. To illustrate, a probability of 100% (0)% was assigned when the forecasted 24-h intensity change was = (=) the RI threshold of 25 kt. Figure 2 indicates that while the scaled version of the RI index had slightly more skill than either SHF5 or GFDI models; none of the objective guidance performed very well in the Atlantic basin. Also, in contrast to the results shown in Fig. 1 the discriminant version was significantly worse than the scaled version and climatology. However, both the scaled and discriminant versions of the RI index performed quite well for the E. Pacific basin with the discriminant version providing the most skillful RI forecast guidance. The only operational model that showed skill relative to climatology was the GFDL model which exhibited a small amount of skill for this basin. Interestingly, the official NHC forecasts (OFCL) exhibited considerable skill in the E. Pacific basin even though the deterministic models themselves exhibited only slight skill or no skill relative to climatology. One reason for this may be the relatively good performance of the RI index in the E. Pacific since the RI index was provided to forecasters in real-time during the 2006 season. It is worth noting that the comparatively poor (good) performance of the RI index and to a certain extent the deterministic models in the Atlantic (E. Pacific) during the 2006 hurricane season can be partly explained by the less (more) than normal activity in the Atlantic (E. Pacific) basins since climatology is a tougher (easier) forecast to beat in less (more) active years. To illustrate, only 6% of the 2006 Atlantic cases rapidly intensified compared to the climatological average of 12% while 16% of the 2006 E. Pacific cases rapidly intensified compared to the climatological mean of 11%.

FY06-07 Achievements

• An updated scaled version of the RI index was tested in real-time at the OAR (NHC) during the 2005 Atlantic and E. Pacific hurricane season
• The performance of the RI index was evaluated for an independent sample of storms from the 2004 and 2005 Hurricane seasons and the results were presented at the 27th Conference on Hurricanes and Tropical Meteorology.
• The new scaled version of the RI index was officially adopted for Operational use in the Atlantic and E. Pacific basins by the NHC in June of 2006
• A new discriminant version of the RI index was run in real-time during the 2006 Atlantic and E. Pacific Hurricane season.
• The RI index performance during the 2006 Atlantic and E. Pacific Hurricane seasons was evaluated and the results were presented at the 2007 Interdepartmental Hurricane Conference.

FY07-08 Milestones:

• Run the RI index and provide output to the NHC in real-time during the 2007 and 2008 Atlantic and E. Pacific hurricane seasons.
• Evaluate RI index performance and modify as necessary to improve future performance
• Submit manuscript describing the updated version of the RI index

References:

Kaplan, J. and M. DeMaria, 2006: Estimating the likelihood of rapid intensification in the Atlantic and E. Pacific basins using SHIPS model data, Preprints 27th Conference on Hurricanes and Tropical Meteorology , Monterey, CA, Amer. Meteor. Soc..

Kaplan, J., and M. DeMaria, 2003: Large-scale characteristics of rapidly intensifying tropical cyclones in the North Atlantic basin. Wea. Forecasting, 18, 1093-1108.

DeMaria, M., M. Manelli, L.K. Shay, J.A. Knaff, and J. Kaplan, 2005: Further improvements to the Statistical Hurricane Intensity Prediction Scheme (SHIPS), Wea. Forecasting, 20, 531-543.

Wilks, D.S., 1995: Statistical Methods in the Atmospheric Sciences, Academic Press, 467pp.